CN107292857A

CN107292857A - Image processing apparatus and method and computer-readable recording medium

Info

Publication number: CN107292857A
Application number: CN201710223092.3A
Authority: CN
Inventors: 中込启太
Original assignee: Canon Inc
Current assignee: Canon Inc
Priority date: 2016-04-13
Filing date: 2017-04-07
Publication date: 2017-10-24
Anticipated expiration: 2037-04-07
Also published as: CN107292857B; US10388018B2; EP3236418B1; US20170301093A1; EP3236418A3; EP3236418A2

Abstract

The present invention provides a kind of image processing apparatus and method and computer-readable recording medium.Described image processing unit includes：Obtaining unit, it is configured to obtain the first image of subject and the second image of the subject；Difference unit, it is configured to obtain the difference image after by described first image and second image registration；And changing unit, it is configured to based on the likelihood score using the calculated for pixel values in the pixel value and second image in described first image, the processing for the pixel value being changed in the difference image.

Description

Image processing apparatus and method and computer-readable recording medium

Technical field

The present invention relates to a kind of image processing apparatus and image processing method and computer-readable recording medium.

Background technology

In medical domain, (CT (Computed are hereinafter referred to as using such as computer tomography device is passed through Tomography, computer tomography) device) etc. the image that obtains of various types of medical science camera devices (mode) enter Row diagnosis.Especially, in order to shoot the situation of subject with the change of time, the image that will be obtained in different timings is compared Compared with.One in the image to be compared will hereinafter be referred to as reference picture picture, and other images will hereinafter be referred to as Deformed article image.

As making the visual image procossing of the change with the time between reference picture picture and deformed article image, it is known that one Plant the technology for the image (will hereinafter be referred to as difference image) for obtaining the difference between representative image, i.e. subtracting techniques. In difference image, the part of the change between two images is depicted, and remaining unchanged part is expressed as having The region of uniform concentration value.

In difference image, due to the difference or subject of imaging conditions between reference picture picture and deformed article image The change of situation, may depict the region in addition to the region that doctor pays close attention on difference image, so as to deteriorate region-of-interest Observability.Japanese Unexamined Patent Publication 2014-94036 publications disclose a kind of based on from one in reference picture picture and deformed article image The noise region that person determines adjusts the technology of the weight of the pixel value on difference image.

However, in technology described in Japanese Unexamined Patent Publication 2014-94036 publications, the region paid close attention to doctor independently, Noise region is determined from one of reference picture picture and deformed article image, and adjusts the power of the pixel value on difference image Weight.Therefore, in the region that the doctor on difference image is not concerned with, change is also depicted on image, so as to deteriorate The observability of difference value in the region of doctor's concern.Further, since only considering reference picture picture when adjusting the weight of pixel value One of with deformed article image, it is therefore desirable to select operation in these images, adjusting the image that weight is based on.

The present invention is made that in view of the above problems, and the present invention provides a kind of pixel that can be used in multiple images Value adjusts the image processing techniques of the weight of the pixel value on difference image.The present invention provides a kind of obtain can be by reality The corresponding arrangement in the pattern of the invention being described later on is applied come function that is realizing but can not being obtained by correlation technique With the technology of effect.

The content of the invention

According to an aspect of the invention, there is provided a kind of image processing apparatus, described image processing unit includes：Obtain Unit, it is configured to obtain the first image of subject and the second image of the subject；Difference unit, it is configured to Obtain the difference image after by described first image and second image registration；And changing unit, it is configured to Based on the likelihood score calculated using the pixel value in the pixel value in described first image and second image, it is changed The processing of pixel value in the difference image.

According to an aspect of the invention, there is provided a kind of image processing apparatus, described image processing unit includes：Obtain Unit, it is configured to obtain the first image of subject and the second image of the subject；Difference unit, it is configured to Obtain the difference image after by described first image and second image registration；And changing unit, it is configured to Between pixel value based on each pixel in described first image and the pixel value of each pixel in second image Compare, the processing for the pixel value being changed in the difference image.

According to an aspect of the invention, there is provided a kind of image processing apparatus, described image processing unit includes：Obtain Unit, it is configured to obtain the first image of subject and the second image of the subject；Difference unit, it is configured to Obtain the difference image after by described first image and second image registration；Display processing unit, it is configured to Described first image and second image are shown on the display unit；And changing unit, it is configured to based on described aobvious Show the display condition of unit, the processing for the pixel value being changed in the difference image.

According to another aspect of the present invention there is provided a kind of image processing method, described image processing method includes：Obtain First image of subject and the second image of the subject；Obtain by described first image and second image registration Difference image afterwards；And based on the calculated for pixel values in the pixel value and second image used in described first image Likelihood score, the processing for the pixel value being changed in the difference image.

According to another aspect of the present invention there is provided a kind of image processing apparatus, described image processing unit includes：Obtain Unit, it is configured to the image for obtaining subject；And characteristic quantity obtaining unit, it is configured to, in described image Concerned pixel, set by multiple predefined paths of the concerned pixel, for each in path, based on the neighbour on path Similarity between nearly pixel value, calculates the successional evaluation of estimate of the pixel value on delegated path, and based on for each road The evaluation of estimate that footpath is obtained, obtains the characteristic quantity of the concerned pixel.

In accordance with the invention it is possible to adjust the weight of the pixel value on difference image using the pixel value in multiple images. In addition, according to the present invention, when adjusting the weight of pixel value, it is not necessary to carry out selection to be used for adjusting the reference picture of weight as Operation, so as to alleviate the burden to the doctor as user.

Description by following (referring to the drawings) to exemplary embodiment, other features of the invention will be clear.

Brief description of the drawings

Fig. 1 is the block diagram for showing to be arranged according to the function of the image processing apparatus of embodiment；

Fig. 2 is the flow chart of the example for the processing procedure for illustrating image processing apparatus；

Fig. 3 is the figure of the display example of image for showing to be obtained by image acquiring unit；

Fig. 4 is to show the likelihood score based on pixel (pixel is included in the pixel in the region of user's concern) (likelihood) come the figure of the example that changes each pixel value in difference image；

Fig. 5 is to show the difference when position skew occurs in the result detected to the correspondence position between image The figure of the example of image；

Fig. 6 be show the image that is obtained by image acquiring unit and each pixel value change after difference image show The figure of example；

Fig. 7 is the flow chart of the example for the processing procedure for illustrating image processing apparatus；

Fig. 8 is the flow chart for illustrating the processing procedure for extracting the region outside region-of-interest；

Fig. 9 is the figure for being used to illustrate to strengthen the example of the processing in block (massive) region；

Figure 10 is the figure for illustrating the example for the processing for strengthening block region；And

Figure 11 A to Figure 11 D are the figures for illustrating the example for the processing for strengthening block region.

Embodiment

Embodiments of the invention are illustratively described in detail now with reference to accompanying drawing.Note, described in embodiment Inscape is only example.The technical scope of the present invention is determined by the scope of claim, and is not implemented by following Example limitation.

According between the image processing apparatus of first embodiment acquisition multiple images (reference picture picture and deformed article image) Difference image, and perform the image procossing of the weight for suitably adjusting the pixel value on difference image.At the image Reason device is characterized in, is obtaining difference image and during performing the processing of image procossing, from the pixel value in reference picture picture and The pixel value in the deformed article image with the deformable registration of reference picture picture is experienced, obtaining pixel, (pixel is included in Pixel in the region of user's concern of doctor etc.) index (likelihood score), and use obtained index (likelihood Degree) it is used as the weight of difference value.

In the art, for example, can be set by the weight for the pixel being included within the region-of-interest on difference image The weight for the pixel being set in the region being not concerned with more than user, to improve the visible of the difference value in the region that user pays close attention to Property.Alternatively, user's concern can be set smaller than by the weight for the pixel being included within the region that user is not concerned with Region in pixel weight, relatively to improve the observability of the difference value in region-of-interest.

Arrangement and the processing of image processing apparatus according to the present embodiment are described below with reference to Fig. 1.Fig. 1 is to show bag Include the frame of the example of the arrangement of the image processing system (magic magiscan) of image processing apparatus according to the present embodiment Figure.Image processing system includes the image processing apparatus 10, network 21 and database 22 as its functional part.Image procossing is filled Put 10 and be communicatively connected to database 22 via network 21.Network 21 includes such as LAN (LAN) and WAN (wide area network).

Database 22 is kept and management medicine image and the information associated with medical image.Image processing apparatus 10 can The medical image kept in database 22 is obtained via network 21.Image processing apparatus 10 includes the (communication of communication IF (interface) 31 Unit), ROM (read-only storage) 32, RAM (random access memory) 33, memory cell 34, operating unit 35, display unit 36 and control unit 37.

Communication IF 31 (communication unit) is formed by LAN card etc., and realize external device (ED) (for example, database 22 etc.) with Communication between image processing apparatus 10.ROM 32 is formed by nonvolatile memory etc., and stores various programs.RAM 33 Formed by volatile memory etc., and the various information of interim storage are used as data.Memory cell 34 is by HDD (hard disk drive) Deng formation, and various information are stored as data.Operating unit 35 is formed by keyboard and mouse, touch panel etc., and will Instruction input from user (for example, doctor) is to various devices.

Display unit 36 is formed by display etc., and shows various information to user.Control unit 37 is by CPU (centers Processing unit) etc. formed, and comprehensively control image processing apparatus 10 in processing.Control unit 37 includes being used as its function Image acquiring unit 50, anamorphose unit 51, difference processing unit 52, the pixel value of part change unit 53 and display processing Unit 54.

In the image processing apparatus according to the present embodiment, image acquiring unit 50 obtain subject the first image and by Second image of a corpse or other object for laboratory examination and chemical testing.First image and the second image are the images in different timing acquisitions.Image acquiring unit 50 is from number The first image I1 (reference picture picture) and the second image I2 (deformed article image) are obtained according to storehouse 22.These images are by various moulds The image (medical image) for the subject that state is obtained.It is in different date and time acquisitions that the present embodiment, which will describe medical image, CT images example.However, medical image can be other kinds of image.How the species of image all can be using this Embodiment.

Anamorphose unit 51 passes through each picture in the position based on each pixel in the first image and the second image Corresponding relation between the position of element makes the second anamorphose so that each pixel in the second image and pair in the first image Pixel matching is answered, by the first image and the second image registration.That is, anamorphose unit 51 is by obtaining the first image I1 (ginsengs According to image) in the position of each pixel and the position of each pixel in the second image I2 (deformed article image) between pair It should be related to, and the second image I2 is deformed based on corresponding relation so that each pixel and the first image in the second image I2 Respective pixel matching in I1, by the first image and the second image registration.Second image I2 deformation process result is below In will be referred to as the image I2' of image I2'(second).In order to calculate corresponding relation, existing linear deformation algorithm can be used, showed Some nonlinear deformation algorithms or combinations thereof.By carrying out the deformable registration between image by anamorphose unit 51, It can will represent to include the characteristic point of the characteristic in the first image I1 and represent to include in second image I2 (the second images I2' the Characteristic points match of the characteristic in).

Difference processing unit 52 obtains the difference image after by the first image and the second image registration.Difference processing list First 52 image after registration obtains the pixel value of corresponding position, and is carried out by the pixel value for being obtained at difference Manage to obtain difference image TS.That is, difference processing unit 52 is obtained at the same position in the first image I1 and the second image I2' Pixel pixel value, difference processing is carried out between the two pixel values, and result of calculation is defeated as difference image TS Go out to pixel value and change unit 53.The pixel subtracted from the pixel value in the first image I1 in the second image I2' is described below The difference processing of value.On the contrary, can use at the difference of the pixel value in the pixel value subtracted image I1 in the second image I2' Reason.

Based on the likelihood score using the calculated for pixel values in the pixel value and the second image in the first image, pixel value changes Unit 53 is changed the processing of the pixel value in difference image.Pixel value changes unit 53 and uses the area with being paid close attention on user The distributed intelligence of pixel value in the associated learning data of the information (pixel value information) of the scope of pixel value in domain (is represented The distributed intelligence of the distribution of pixel value in region-of-interest), based on being calculated seemingly from the first image I1 and the second deformation pattern I2' Right degree, come the processing of the pixel value of each pixel being changed in difference image TS.

The information (pixel value information) of the scope for the pixel value paid close attention on user, can be by user via operating unit 35 Input, or the first image I1 and/or the second image I2 that are shown from display unit 36 are automatically determined.For example, pixel value Change unit 53 to obtain in each pixel in difference image on closing based on the information inputted via operating unit 35 The pixel value information of the scope of pixel value in note region.In addition, pixel value change unit 53 can be based on display unit 36 The pixel value letter on the scope of the pixel value in region-of-interest is obtained in display condition, each pixel in difference image Breath.Pixel value change unit 53 can the display condition based at least one of the first image and the second image, to be closed The pixel value information of the scope of pixel value in region-of-interest.

Pixel value changes unit 53 based on the pixel value information obtained, to set the pixel value paid close attention in region The distributed intelligence of distribution.The distributed intelligence of pixel value in the storage learning data of memory cell 34 (pays close attention to the picture in region Element value distribution distributed intelligence), and pixel value change unit 53 be based on obtained according to pixel value information from memory cell 34 Distributed intelligence distributed intelligence is set.The distributed intelligence of pixel value in learning data (pays close attention to the pixel in region The distributed intelligence of the distribution of value) correspond to each position (for example, lung, bone, liver etc.) of subject, and it is stored in storage The information for the different distributions for representing pixel value is used as in unit 34.Pixel value information in region-of-interest and the picture in learning data The distributed intelligence (distributed intelligence for paying close attention to the distribution of the pixel value in region) of plain value is associated.If for example, pixel value Change unit 53 obtains the pixel value information on the bone region as region-of-interest, then is obtained and bone region from memory cell 34 (distribution for paying close attention to the distribution of the pixel value in region is believed for the distributed intelligence of pixel value corresponding, in learning data Breath), and the distributed intelligence is set to likelihood score design conditions (θ).Pixel value changes unit 53 and is based on using the first image In pixel value and the calculated for pixel values in the second image likelihood score, come the place of pixel value being changed in difference image Reason.

Note, the distributed intelligence of the pixel value in learning data (pays close attention to the distribution of the distribution of pixel value in region Information) it is the information resulted in when various mode obtain image.The distributed intelligence of pixel value in learning data (is represented The distributed intelligence of the distribution of pixel value in region-of-interest) number can be stored in together with the image (medical image) of subject According in storehouse 22.In this case, point of the pixel value during image acquiring unit 50 can obtain learning data from database 22 Cloth information (distributed intelligence for paying close attention to the distribution of pixel value in region), and store it in memory cell 34.

The result of the processing for the pixel value being changed in difference image TS is exported next, pixel value changes unit 53 To display processing unit 54.The difference image for performing pixel value change processing based on likelihood score will hereinafter be referred to as difference Image TS'.

Display processing unit 54 is used as the display control unit of the display for control display unit 36.Display processing unit 54 show the difference image TS' calculated by pixel value change unit 53 in the image display area of display unit 36.At display Reason unit 54 can show the first image all obtained by image acquiring unit 50 in the image display area of display unit 36 I1 and the second image I2, the second image I2' deformed by anamorphose unit 51 and calculated by difference processing unit 52 Difference image TS.For example, display processing unit 54 is capable of the display of control display unit 36, to be displayed side by side difference image TS' With TS, the first image I1 and the second image I2 (I2'), or it is superimposed and shows some in these images.

Each part of image processing apparatus 10 is operated according to computer program.For example, control unit 37 (CPU) is used RAM 33 is carried in the computer program stored in ROM 32 or memory cell 34 as working region, and performs them, So as to realize the function of each part.Note, special circuit can be used realize image processing apparatus 10 part some or Repertoire.Alternatively, cloud computer can be used come some functions of the part of realizing control unit 37.

For example, the operation device present in the place different from the place of image processing apparatus 10 can be via network 21 It is communicatively connected to image processing apparatus 10.It is then possible to by between image processing apparatus 10 and operation device send/ Data are received, come the function of the part of realizing image processing apparatus 10 or control unit 37.

Referring next to Fig. 2 to and Fig. 6 the example of the processing of the image processing apparatus 10 shown in Fig. 1 described.Fig. 2 It is the flow chart of the example for the processing procedure for illustrating image processing apparatus 10.The present embodiment, which will be enumerated, each includes the medical science figure of bone Picture.However, the present embodiment may be used on each including the medical image of other concerns position (such as lung, brain or liver).

(step S101：Acquisition/display of image)

In step S101, if user indicates to obtain reference picture picture (the first image I1) and deformation via operating unit 35 Object images (the second image I2), then image acquiring unit 50 multiple images (the reference specified by user is obtained from database 22 Image (the first image I1) and deformed article image (the second image I2)), and be stored in RAM33.In addition, as schemed Shown in 3, display processing unit 54 shown in the image display area 300 of display unit 36 from database 22 obtain it is multiple Image (reference picture picture (the first image I1) and deformed article image (the second image I2)).

(step S102：Deformable registration, step S103：Anamorphose)

In step s 102, anamorphose unit 51 reads image from RAM 33, and calculates each in the first image I1 The corresponding relation between each pixel in individual pixel and the second image I2.Represented more specifically, anamorphose unit 51 is calculated From each pixel in the first image I1 to the deformation vector of the corresponding relation of each pixel in the second image I2.Deformation vector It is to represent in pair, deformed article image (the second image I2) corresponding with each pixel in reference picture picture (the first image I1) Each pixel virtual amount of movement (displacement) and the vector of moving direction (deformation direction).For example, in the feelings of 3-D view Under condition, if the coordinate (x1, y1, z1) of each pixel in reference picture picture (the first image I1) is moved to deformed article image The coordinate (x2, y2, z2) of each pixel in (the second image I2), then represent deformation vector by (x2-x1, y2-y1, z2-z1).

Note, the linear image deformable registration method of affine transformation etc., such as LDDMM (Large can be passed through Deformation Diffeomorphic Metric Mapping, large deformation differomorphism measures mapping) etc. non-linear figure As deformable registration method or combinations thereof, to perform the calculating of the deformation vector between the correspondence position in image.By scheming As the multiple images of the acquisition of obtaining unit 50 can be the image in addition to the original image obtained by various mode.For example, The output image of the various image enhancing filters of edge enhancement filter etc. can be used, by extracting region-of-interest and The area image of acquisition, and these images combination.

In step s 103, anamorphose unit 51 is using the deformation vector obtained in step s 102, according to the second figure As I2 generates the second image I2' so that in each pixel and deformed article image (the second image I2) in the first image I1 Respective pixel is matched, and the image of generation is stored in RAM 33.Image of the display processing unit 54 in display unit 36 The the second image I2' generated by anamorphose unit 51 is shown in viewing area.

Note, if the position (pixel) for being included in the first image I1 and the subject in the second image I2 is initially right each other Should, then it can skip the processing in step S102 and S103.

(step S104：Difference processing (generation of difference image) between pixel value)

In step S104, difference processing unit 52 reads the first image I1 and the second image I2' from RAM 33, by entering Difference processing generation difference diagram between the pixel value of the pixel of row the first image I1 and the corresponding position in the second image I2' As TS, and the image of generation is stored in RAM 33.

(step S105：Obtain the pixel value information in the region of user's concern)

In step S105, pixel value change unit 53 obtain on user pay close attention to pixel value scope information (as Plain value information).Pixel value changes display condition of the unit 53 based on the image shown on display unit 36, obtains and is closed on user The information (pixel value information) of the scope of pixel value in the region of note.It is based on more specifically, pixel value changes unit 53 in step The display bar for the image (the first image I1 and the second image I2 or I2') being shown in rapid S101 or S103 on display unit 36 Part, for example, such as window position (the window level, WL) and window width (window width, WW) etc. that are changed for concentration value Arranges value, the pixel value information in region to estimate user's concern.

When carrying out the diagnostic imaging of CT images as the doctor of user, he/her passes through the subject according to doctor's concern Position change window position (window level, WL) and window width (window width, WW), to change and doctor concern is detected The setting (concentration value of conversion display image) of the display condition of the corresponding image in the position of body.If more specifically, for example, Doctor carries out the diagnostic imaging of bone, then window position (WL) is set into value between 350 to 500 [H.U.], and by window width (WW) It is set to the value between 1500 to 2500 [H.U.].By the setting for the display condition for changing image according to the position of subject, The display image of the concentration distribution at position (for example, bone) of the display with the subject for making it easy to watch doctor's concern.

By using this point, pixel value changes unit 53 being capable of image (the first figure based on display experience difference processing As I1 and the second image I2 or I2') display condition for images (for example, window position (WL) and window width (WW)) of display unit 36 set Put value, come estimate as user doctor concern pixel value scope and/or the position (organ) of subject.

Note, if window position (WL) and window width (WW) be arranged on it is different between the first image I1 and the second image I2', Expect using the benchmark of deformable registration, the arranges value of reference picture picture (the first image I1) served as between image, but can To use the second image I2' arranges value.Pixel value change unit 53 can based in the first image and the second image at least The display condition of one, to obtain the pixel value information on the scope of the pixel value in region-of-interest.

The pixel value information of user's concern can be inputted from user via operating unit 35.Each picture in difference image In element, pixel value changes unit 53 can be based on the information inputted via operating unit 35, to obtain the picture on region-of-interest The pixel value information of the scope of element value.Pixel value change unit 53 can be obtained from storage destination is stored in ROM 32 or RAM The pixel value information of predetermined (predetermined) in 33.

(step S106：Change the pixel value (image subtraction operation result) in difference image)

In step s 106, pixel value change unit 53 uses the study number associated with the pixel value information that user pays close attention to The distributed intelligence (distributed intelligence for paying close attention to the distribution of the pixel value in region) of pixel value in is calculated as likelihood score Condition (θ), based on the likelihood score according to the first image I1 and the second image I2' calculating, to change each in difference image TS The pixel value of pixel, so as to generate difference image TS'.

It is the information being expressed as follows according to the first image I1 and the second image I2' likelihood scores calculated：With difference image TS In position is corresponding, the pixel at the first image I1 and/or position (same coordinate) place in the second image I2 (I2'), quilt It is included in the possibility in the region of user's concern.

In order to calculate the likelihood score of pixel value, it is assumed that the distribution of pixel value is normal distribution, and pixel value changes unit 53 calculate the likelihood score of pixel value using the distributed intelligence of the distribution for the pixel value paid close attention in region.Pixel value changes single Member 53 from known learning data, be obtained ahead of time as the pixel value paid close attention in region distribution it is distributed intelligence, just The parameter (average value and variance) of state distribution, and by using the distributed intelligence (parameter of normal distribution) obtained as seemingly Design conditions (θ) are so spent, to calculate the likelihood score of each pixel in image to be processed.Therefore, carrying out likelihood score calculating Before, pixel value changes the distributed intelligence that unit 53 obtains the distribution as the pixel value in paying close attention to region from learning data , the parameter of normal distribution.

For example, in the case of CT images, pixel value changes unit 53 and is directed to each portion existed with each concentration range Position (organ) (for example, the abdomen organ of the lung areas comprising large quantity of air including the liver being mainly made up of soft tissue or Person's bone), obtain distributed intelligence, normal distribution the parameter of the distribution as the pixel value paid close attention in region.

In being calculated according to the likelihood score of the present embodiment, pixel value changes unit 53 by the pixel in the region paid close attention to user The distributed intelligence of pixel value in the associated learning data of value information (pays close attention to the distribution of the distribution of pixel value in region Information) likelihood score design conditions (θ) are set to, and based on the pixel in the pixel value and the second image used in the first image The likelihood score that value is calculated, the processing for the pixel value being changed in difference image.Make more specifically, pixel value changes unit 53 With weight of the likelihood score as the pixel value (difference value) in difference image.Pixel value changes unit 53 and used based on likelihood score institute The weight coefficient of acquisition changes the pixel value in difference image.Pixel value changes the pixel value that unit 53 will be paid close attention to including user Region in difference value be set to comparatively be more than remaining area in difference value.This can only strengthen user and want seizure Change, so as to improve the observability of difference image.

Fig. 4 shows the image display example when region paid close attention to as user is bone.Reference picture 4, difference image TS 403 Be based on the reference picture shown in Fig. 4 as 401 (the first image I1) and deformed article image 402 (the second image I2) therebetween Difference and the image that obtains, and difference image TS'404 is by performing the difference diagram that pixel value changes processing and obtained Picture.In Fig. 4 difference image TS and TS', principal-employment score value, minus tolerance score value and difference are represented by white, black and grey respectively Value 0.Dotted line represents the profile of area (organic region).However, area (organic region) is necessarily depicted in difference On image.Fig. 4 example is shown when the learning data in the bone region from the area paid close attention to as user obtains expression bone Situation during distributed intelligence (parameter) of the distribution of the pixel value in region.In this case, based on the use distributed intelligence Difference between the likelihood score that (parameter) is calculated, the pixel that will be likely included in the area (bone region) of user's concern The weight of value is set greater than the weight of the difference value between the pixel in remaining area, and changes pixel value.As described above, Can by enhancing be likely included in user pay close attention to area (bone region) in pixel between difference, suppress except The description of the change with the time in region (for example, abdomen organ) beyond the area of user's concern.

In the present embodiment, pixel value changes unit 53 based on being obtained seemingly according to the first image I1 and the second image I2' Right degree, come the weight coefficient W (p) of the multiplication that defines the pixel value that be used to change the pixel p in difference image TS, as following Given by formula (1).Pixel value changes unit 53 based on the pixel value (I1 (p)) used in the first image and pays close attention to region In pixel value distribution the likelihood score Pr (I1 (p) | θ) that is calculated of distributed intelligence (θ) and using the picture in the second image Plain value (I2'(p)) and pay close attention to pixel value in region distribution the likelihood score Pr (I2' that are calculated of distributed intelligence (θ) (p) | θ)) in the greater, come the processing of pixel value being changed in difference image.

W (p)=max (Pr (I1 (p) | θ), Pr (I2'(p) | θ)) ... (1)

Wherein, function max (A, B) is the function for extracting from the maximum in variables A and B, and Pr (C | D) represent when to shaping The probability (likelihood score) of condition C is obtained during part D.In addition, I1 (p) represents the pixel value of the pixel p in the first image I1, I2'(p) Represent the pixel value of the pixel p in the second image I2'.If the second image I2 is not deformed, I2'(p) represent the second image Pixel value in I2.

Moreover, it is assumed that θ represent being obtained from learning data, pay close attention to region in pixel value distribution distributed intelligence (parameter).Then, pixel value changes each in all pixels that unit 53 is directed in difference image TS, by by difference image Pixel value TS (p) in TS is multiplied by weight coefficient W (p) to change pixel value, and will be as by performing at pixel value change The difference image TS' for managing the result obtained is stored in RAM 33.

Both images (the first image I1 and the second image I2') are used in the formula (1) for calculating weight coefficient W (p) As difference image TS calculating source the reason for be, in order to prevent mistakenly suppressing the difference in the region-of-interest on difference image Value.

If for example calculating weight coefficient W (p) using only one of image, if there is having in one image There is high likelihood score but there is the pixel of low likelihood score in another image, then weight may diminish.If more specifically, with to The difference with the time between the image that timing section is shot is object, then has in one image and represent normal organ structure The pixel of concentration value, may have the pixel value outside the scope of normal pixel value in another image due to lesion. Therefore, in this case, if diminished using the image including lesion to calculate the weight at weight, respective pixel.

If on the contrary, calculate weight coefficient W (p) both as difference image TS calculating source using image, if Pixel value falls in the range of the likelihood score of in the picture at least one, then the change of the pixel in the region that can be paid close attention to user (difference value) provides big weight.If in addition, using both images, when calculating weight coefficient W (p), it is not necessary to select The step of one in first image I1 and the second image I2'.

In Fig. 4 in shown example, in the difference as the simple differencing between the first image I1 and the second image I2' In image TS, in addition to the area (bone region) in the region paid close attention to as user, the region being not concerned with as user Other areas (liver area) also serve as difference value reservation.CT values in liver area generally fall in 60 to 70 [H.U.'s] In the range of, and CT values in bone region are equal to or more than 150 [H.U.].Therefore, by learn the expression in learning data with Pr (I1 (p) | θ) and Pr (I2'(p) in distributed intelligence (parameter) θ of the distribution of the relevant pixel value in bone region, bone region | θ) Higher than the Pr (I1 (p) | θ) and Pr (I2'(p) in liver area | θ).Therefore, by the pixel value TS (p) in difference image TS It is multiplied by weight coefficient W (p) result (that is, TS'(p)) in, the pixel value TS'(p in other areas (liver area)) with Pixel value TS'(p in the area (bone region) in the region paid close attention to as user) compare, with the difference value close to 0.Knot Really, the difference value in the difference image TS' as shown in Fig. 4, other areas (liver area) is suppressed, and is being made The relative enhancing of difference value in the area (bone region) in the region paid close attention to for user, so that being easier visually to perceive It is used as the change of the bone of region-of-interest.

Note having when calculating weight coefficient W (p) using image both as the calculating source of difference image following excellent Point.That is, as difference image TS feature, the meter of the correspondence position in the image that is carried out in step S102 and S103 can be made Error in calculation is clearly visualized.This feature is described in detail in reference picture 5.Note, in Figure 5 in shown difference image, Similar to Fig. 4, principal-employment score value, minus tolerance score value and difference value 0 are represented by white, black and grey respectively.Assuming that user's concern Region is the bone region in Fig. 5.

Reference picture 5, image 501 is reference picture picture (the first image I1), and image 502 is deformed article image (the second image ), and image 503 is deformed article image (the second image I2') I2.Image 504 shows (the reference picture picture (first of image 501 Image I1)) and image 502 (deformed article image (the second image I2)) superimposed image.Image 505 shows the (ginseng of image 501 According to image (the first image I1)) and deformed article image (the second image I2') superimposed image.

Image 506 is illustrated as use image (the first image I1 and the second image I2') both as difference image Calculating source calculates weight coefficient and changes each pixel value in difference image by the way that pixel value is multiplied by into weight coefficient Result difference image TS'.

It is small for occurring in the detection of correspondence position in the picture as shown in Fig. 5 image 506 (difference image) The place of position skew, it is close to each other on the occasion of being depicted as with negative value on difference image.The difference on image is seen as user The rendering results of value, for example, positive pixel value and negative pixel values locally inverted place when, he/her can recognize pixel value The place that is really changed and the wrong place that position skew is there occurs in correspondence position testing result between images.

Image 507 is illustrated as using one of image (the first image I1) to be counted as the calculating source of difference image Calculate weight coefficient and change the result of each pixel value in difference image by the way that pixel value is multiplied by into weight coefficient Difference image TS'.Image 508 is illustrated as using one of image (the second image I2') as the meter of difference image Calculation source calculates weight coefficient and changes each pixel value in difference image by the way that pixel value is multiplied by into weight coefficient The difference image TS' of result.

If as described above, position skew occurs in correspondence position testing result between images, then it is assumed that for the Pixel in the pixel at same coordinate point in one image I1 and the second image I2', an image has high bone likelihood score, And the pixel in another image has low bone likelihood score.Therefore, if using only the likelihood calculated according to one of image Spend to determine the weight of difference value, then as shown in image 507 and image 508, the weight of one in difference value is undesirably It is small.Therefore, exist close to each other on difference image on the occasion of the Character losing with negative value, therefore, it is difficult to recognize pixel Place and the wrong place of correspondence position testing result that value has veritably been changed.

On the other hand, if calculating power using both images used when calculating difference value (Fig. 5 image 506) Weight, then keep on the occasion of the feature existed close to each other with negative value.Therefore, using both images (the first image I1 and the second figure As I2') weight coefficient is calculated as difference image TS calculating source there is following effect：As Fig. 5 image 506 in institute Show, maintain the feature of the easy wrong identification of difference image TS correspondence position testing result.

(step S107：The output display of result of calculation)

In step s 107, difference image TS' is output to display processing unit 54 by pixel value change unit 53.Such as Fig. 6 Shown in, display processing unit 54 shows difference image TS' in the image display area 300 of display unit 36.Display processing The display of the control display unit 36 of unit 54 be displayed side by side difference image TS', reference picture picture (the first image I1) and deformation pair As image (the second image I2).Display processing unit 54 is capable of the display of control display unit 36 further to show the second image (I2')。

, can be based on the scope of the pixel value with being paid close attention on user in the image processing techniques according to the present embodiment The distributed intelligence of pixel value in the associated learning data of information (pixel value information) (pays close attention to pixel value in region The distributed intelligence of distribution) and reference picture as and deformed article image both, to adjust each pixel value on difference image Weight.The processing can strengthen the description of the change for changing and suppressing non-interesting position at concern position, so as to improve user The observability of the change at the position of concern.

(modified example 1)

Although in step s 106, pixel value change unit 53 uses the pixel p in the first image I1 and the second image I2' Pixel value calculate the weight coefficient at pixel p, but the pixel value of the neighborhood pixels of pixel p can be used (for example, being directed to 6 or 26 neighborhood pixels of 3-D view) calculate.It can be based on using in the first image for example, pixel value changes unit 53 Pixel and its likelihood that is calculated of the pixel value of neighborhood pixels and the pixel in the second image and its pixel value of neighborhood pixels Degree, come the processing of pixel value being changed in difference image.In the calculating for obtaining likelihood score, the pixel for assigning pixel is used Value carrys out defined formula (formula (1)).The image enhancing filter of smoothing filter or edge enhancement filter etc. can be used Output valve be used as the value.

In addition, if it is possible to carried by the known method for extracting region of threshold process or pattern cut segmentation etc. The pixel in region-of-interest is taken, then probability 1 is assigned to the region extracted, and the region of the region exterior to being extracted is assigned Probability 0, weight coefficient W is used as thereby using probability.

Note, pixel value changes the minimum value and maximum that unit 53 can be based on the weight coefficient W (p) calculated, come By weight coefficient W normalization in preset range (for example, [0,1]).

(modified example 2：The modified example 1 of the definition of weight)

In first embodiment and modified example 1, by using two images (the first image I1 and the second figures based on formula (1) As I2') in pixel value obtain likelihood score, to determine the weight coefficient W for changing the pixel value TS (p) in difference image TS (p).As another method, as provided by following formula (2), for by the pixel value in two images to defined point Cloth obtains likelihood score.In modified example 2, pixel value changes unit 53 can be based on using the pixel value paid close attention in region Distribution distributed intelligence (θ) and the pixel value obtained is calculated based on the pixel value in the first image and the second image Likelihood score, come the processing of pixel value being changed in difference image.

W (p)=Pr (I1 (p), I2'(p) | θ) ... (2)

Although using the likelihood score based on single argument normal distribution in formula (1), using based on double changes in formula (2) Measure the likelihood score of normal distribution.Similar to the situation using single argument normal distribution, equally in multivariable (two or more Variable) in normal distribution, pixel value changes unit 53 also can be by obtaining parameter (average value of variable and side from learning data Difference-covariance matrix) as the distributed intelligence (θ) for the distribution for representing pixel value, to calculate likelihood score.Expression picture in formula (2) The distributed intelligence (θ) of the distribution of element value includes average value, the second image of the pixel value in the region-of-interest in the first image I1 The average value and the variance-covariance matrix of value of the pixel value in region-of-interest in I2'.Pixel value changes unit 53 can The distributed intelligence that is obtained using the distributed intelligence that obtains when shooting the first image and when shooting the second image calculates likelihood Spend the distributed intelligence (θ) as the distribution for representing pixel value.In this case, pixel value changes unit 53 for the first image Pixel value in I1, using the distributed intelligence obtained when shooting the first image, and for the picture in the second image I2 (I2') Element value, using the distributed intelligence obtained when shooting the second image, to calculate likelihood score.Note, pixel value changes the energy of unit 53 One of enough distributed intelligences obtained using the distributed intelligence obtained when shooting the first image and when shooting the second image, To calculate the likelihood score for the pixel value in the first image and the second image.

As the method being distributed using multivariate normal, except the picture of the pixel p in the first image I1 and the second image I2' Beyond element value, additionally it is possible to defined by combining the output result of the image processing filter for the first image and the second image Likelihood score.In addition it is possible to use the output knot of the neighborhood pixels value of pixel p and the neighborhood pixels for passing through image processing filter Really.

Note, minimum value and maximum that can be based on the value calculated, by the way that weight coefficient W (p) is normalized pre- Determine in scope (for example, [0,1]), to obtain weight coefficient W.

(modified example 3：The modified example 2 of the definition of weight)

In first embodiment and modified example 1 and 2, the likelihood score based on learning data is defined, and determines to be used to change The weight coefficient W (p) of pixel value (p) in difference image TS.However, as given by following formula (3), image can be used In pixel value be used as weight coefficient.In modified example 3, pixel value changes unit 53 based on each pixel in the first image Comparison between pixel value and the pixel value of each pixel in the second image, come the pixel value that is changed in difference image Processing.Pixel value changes unit 53 and changes the pixel in difference image TS using based on comparing the weight coefficient W (p) of acquisition Value.

W (p)=max (I1 (p), I2'(p)) ... (3)

As can applying equation (3) example, user concern region in pixel value comparatively be more than other regions in Pixel value.More specifically, for example it is to be noted that bone region in CT images.There is the X-ray than other regions to inhale in known bone region Comparatively pixel value in the higher X-ray absorption rate of yield, therefore bone region in CT images is more than in other regions Pixel value.Therefore, it is possible to by setting big weight to the big region of the pixel value in the first image I1 or the second image I2', So as to compared with other regions, strengthen the difference value in the bone region in difference image TS.

Note, expect weight coefficient W being calculated as 0 or bigger value.Minimum value that can be based on the value calculated and Maximum, by weight coefficient W (p) normalization in preset range (for example, [0,1]).

In this example, weight coefficient W (p) is provided so that weight becomes bigger as pixel value is more big.However, If for example, noting the region of the lung in CT images etc., can set bigger by weight as pixel value is smaller. More specifically, for example, cause minimum value to be 1 by correction pixels value, obtain that its is reciprocal, and be used as weight coefficient W (p), weight becomes bigger as pixel value is smaller.

Note, in the calculating of formula (3), have been described that the calculating using the pixel value for assigning pixel.However, it is possible to make Pixel value is used as with the output valve of smoothing filter or edge enhancement filter.Alternatively, the neighbouring picture of pixel p can be used The output result of element value and the neighborhood pixels for passing through image processing filter.

(modified example 4：The modified example 3 of the definition of weight)

As from that according to the first embodiment method that sets weight coefficient W different with modified example 1 to 3, can make as follows With the arranges value of the display condition of the image shown on display unit 36, such as it is shown in step S101 on display unit 36 Image window position (WL) and window width (WW).Display condition includes representing the arranges value of the intermediate value of pixel value range and represents relative In the arranges value of the width of the pixel value range of intermediate value.In modified example 4, display processing unit 54 is shown on display unit 36 First image and the second image, and pixel value change unit 53 is changed difference based on the display condition of display unit 36 The processing of pixel value in image.Pixel value changes unit 53 by the weight coefficient that is obtained based on display condition, to change Pixel value in difference image.Display condition includes representing the arranges value of the intermediate value of pixel value range and represented relative to intermediate value The arranges value of the width of pixel value range.

In general, window position (WL) and window width (WW) based on the display condition for being arranged to CT images are determined by from black The scope of pixel value on color to the picture of gray scale (tone) expression of white.Window position (WL) is the picture for representing to be reached by gray scale chart The arranges value of the intermediate value of element value scope, and window width (WW) is the setting of the width of pixel value range for representing to be reached by gray scale chart Value.That is, on picture, (WL-WW/2) or smaller pixel value are expressed by black, and expressed (WL+WW/2) by white Or bigger pixel value.That is, due to falling the pixel value quilt outside the pixel value range specified by window position (WL) and window width (WW) Trimming, even if therefore pixel value in original image is different, they are also expressed by identical black or white on picture.

By using this point, pixel value change unit 53 using the arranges value based on the intermediate value for representing pixel value range and The weight coefficient that is obtained of arranges value of the width of pixel value range is represented, come the place of pixel value being changed in difference image Reason.If the pixel value in the first image I1 or the second image I2' falls in the range of [WL-WW/2, WL+WW/2], then it will add Weight coefficient W is set to 1；Otherwise, weight coefficient W is set to 0.Therefore, it is possible to the scope for the pixel value for only extracting user's concern Interior difference.Note, if window position (WL) and window width (WW) be arranged on it is different between the first image I1 and the second image I2', Then pixel value, which changes unit 53, can use benchmark, the first image I1 the arranges value served as in deformable image registration.Note Meaning, pixel value, which changes unit 53, can also use window position (WL) and the arranges value of window width (WW) for the second image I2'.As Selection, pixel value change unit 53 can use pixel value range and the second image I2' concern of the first image I1 concern Pixel value range between overlapping part pixel value range.

<Second embodiment>(pixel outside region-of-interest is extracted, and it is used for change pixel in difference result Value.)

In the first embodiment, pixel value change unit uses the distribution of the pixel value in region-of-interest, extracts concern area Result of pixel in domain etc., to change the pixel value in difference result.On the other hand, the picture outside region-of-interest can be used The distribution of plain value and extraction result change the pixel value in difference result.

The present embodiment will illustrate for example following situation：If bone is set into region-of-interest, is extracted or strengthened The information in the region of heart or liver outside region-of-interest etc., and the pixel in difference result is changed based on the information Value.

In addition to changing the addition function of unit 53 to pixel value, according to the arrangement of the image processing apparatus of second embodiment It is identical with the arrangement according to first embodiment.In a second embodiment, as shown in Figure 7, the region of user's concern is being obtained After pixel value information, pixel value change unit 53 obtain following information, in the information, from input image data group enhancing or The region that user is not concerned with is extracted, and the pixel value in difference image is changed based on the information.

Reference picture 7 to Figure 10 and Figure 11 A to Figure 11 D is described according to the image processing apparatus 10 of second embodiment Processing procedure.

Fig. 7 is illustrated since the data in image processing apparatus 10 obtain processing, and shows change pixel value The flow chart of the example of the processing of difference image.Among step in the flow chart, in step S1010 to S1050 and In S1070, carry out and the processing identical in the step S101 to S105 and S107 according to first embodiment shown in Fig. 2 Processing.That is, the processing in step S1055 and S1060 is different from first embodiment.The processing of addition will only be described below and with the The difference of one embodiment.

(step S1055：Extract the region outside region-of-interest)

In step S1055, pixel value changes unit 53 from reference picture as being carried with both deformed article images after deformation Take the region that family is not concerned with.For example, if region-of-interest is bone, the organic region of heart and liver etc. is concern area The region in overseas portion.Using the characteristic of pixel value information, for example, by the processing procedure shown in Fig. 8, to extract region-of-interest These outside regions.

As shown in Figure 4, when being compared with bone structure, the organ structure of liver has following characteristic：In the presence of with phase Like the set of the pixel of pixel value to form block (mass).On the contrary, for bone structure, each bone generally has small, long and narrow Structure, and the surface (cortex bone) of bone and the pixel value changes of inner side (cancellous bone) are very big.Therefore, it is possible to by strengthening by having There is the block region of the pixel groups formation of similar pixel value, and by the known extracting method of threshold process etc. from image The region is extracted, to extract the region in addition to bone region.

The separated viewpoint in region by bone region and in addition to bone region from CT images, will illustrate to be based on picture in addition The effect of the characteristic quantity in the continuity enhancing block region of element value.On CT images, bone region has about 150 [H.U.] or bigger High level.The organic region of heart and liver etc. has about 60 [H.U.] or bigger pixel value, but these pixel values can It can be uprised by inspection method (mainly by using contrast agent).Therefore, according to radiography imaging conditions, especially such as heart There is the distribution of the pixel value overlapping with the distribution of the pixel value in bone region with the organic region of liver etc..Therefore, at some In the case of, for the method for extracting region of the threshold process of pixel value itself etc., it is impossible to which reply images bar in various radiographies Extracted region under part.On the other hand, if the successional characteristic quantity based on pixel value is calculated, even if radiography imaging conditions Difference, also results in constant calculated value, and part unrelated with radiography shooting bar, as long as being used as the organic region quilt of extracting object Level dyeing.Therefore, it is used to strengthen the spy of block region rather than pixel value by using in the extraction of liver or heart The amount of levying, can be expected for the difference of radiography imaging conditions to obtain stable extracted region result.

In fig. 8 in shown processing procedure, in step s 201, the image of subject is obtained.In step S202, pin Image is smoothed, so as to reduce the noise in image.Note, can skip in step S202 be used for reduce noise The processing of purpose.Afterwards, in step S203, the parameter (threshold value) for evaluating the similarity between pixel value is obtained.Can Using parameter (threshold value) successional characteristic quantity is represented to calculate.More specifically, predefined parameter can be used, or can by with Family is via the input parameter of operating unit 35.Alternatively, the variance of pixel value that can be in image automatically determines parameter.Such as Fruit automatically determines parameter by this way, then expects to set bigger by the value of parameter (threshold value) as variance is bigger.In step In rapid S204, the successional characteristic quantity for representing pixel value is calculated, and strengthen the block region that image includes.Pixel value changes Unit 53 is directed to the concerned pixel in image, multiple predefined paths by concerned pixel is set, based on the neighbouring picture on path Similarity between element value, calculates the successional evaluation of estimate of the pixel value on each path of representative in each path, and The characteristic quantity (acquisition of characteristic quantity) of concerned pixel is obtained based on the evaluation of estimate obtained for each path.Pixel value changes unit 53 calculate the likelihood score that concerned pixel is in block region also based on the characteristic quantity obtained.In step S205, pixel value Change the extraction unit that unit 53 is used as being used for extracting region from image, and use such as threshold process or pattern cut point The known extracting method of section etc. extracts the block region as the region in addition to bone region.

As the method in block region is strengthened in step S204, for example, it is contemplated that following method：When giving in attention image When fixation is plain, the quantity of the similar pixel of continued presence on the predefined paths through given pixel is calculated.In this example, in advance The quantity for determining path is one or more, strengthens block region it is desirable to multiple paths.Especially, at least pin is expected Path is set to each axial direction of image.I.e. it is desired to when image is three-dimensional, at least for each in three axial directions Individual setting path, and when image is two dimension, at least for each setting path in two axial directions.

Note, in order to calculate the successional characteristic quantity on pixel value, can in advance limit and be arranged to calculate object Pixel (concerned pixel).More specifically, it is considered to following method：There is provided threshold value for pixel value, and be equal to for having or Pixel more than the pixel value (or the pixel value equal to or less than threshold value or pixel value less than threshold value) of threshold value calculates feature Amount.Consider following situation：If the bone in CT images is arranged to pay close attention to region, extracts or strengthen outside region-of-interest The region of heart and liver etc..In the example according to the present embodiment, it is not necessary to for not in organic region The pixel for the pixel value (for example, -100 [H.U.] or smaller) that pixel is kept carries out following calculating., can by skipping calculating Accelerate following characteristic quantity calculates processing.Especially, the model for the pixel value obtained in step S1050 and user paid close attention to It is effective that pixel in enclosing, which is set to calculate object (concerned pixel),.

Reference picture 9 to Figure 10 and Figure 11 A to Figure 11 D is described to strengthen the example of the processing in block region.

To describe it is shown in fig .9 in the case of calculate the successional method of concentration value.Fig. 9 is shown to be surrounded by thick line The set of pixel represent predefined paths, and concerned pixel is the situation of the pixel at x coordinate and y-coordinate (2,3) place.Note, The numerical value described in each shown pixel represents the pixel value of pixel in fig .9.In this case, will be described below as Under method：Calculate the quantity of the similar pixel of continued presence on the predefined paths by concerned pixel.

As shown in Figure 10, when calculating characteristic quantity, start to calculate with concerned pixel (2,3).First, by concerned pixel The pixel value of (2,3) is compared with the pixel value of neighborhood pixels (3,3).If the poor absolute value between pixel value is less than The parameter (threshold value) obtained in step S203, it is determined that pixel value is continuous, and calculating proceeds to next step.In Figure 10 Example in, the pixel (2,3) of the object in being calculated as first and the pixel value of (3,3) are 100 and 110, and pixel value Between poor absolute value be 10.If the threshold value obtained in step S203 is 20, the comparative result of pixel value is less than threshold Value, it is thus determined that pixel (2,3) and (3,3) have continuity.In this case, in the stage, will as pixel (2,3) and The distance between (3,3), the value of the width of pixel, interim storage is successional characteristic quantity (evaluation of estimate).Afterwards, only Want the continuity of pixel to continue, then just slide the pixel pair as calculating object, and repeat calculating.As actually showing Example, will describe to determine that there is pixel next calculating after continuity to walk in the calculating for pixel to (2,3) and (3,3) Suddenly.In next calculation procedure in this case, pixel is slided to so that pixel (3,3) and pixel (4,3) are set into object, And to be compared their pixel value with aforesaid way identical mode.In second calculates, the pixel for calculating object is used as The pixel value of (3,3) and (4,3) is 0 all for the poor absolute value between 110, and pixel value.Therefore, because absolute value is less than As the 20 of the threshold value obtained in step S203, it is thus determined that pixel (3,3) and (4,3) have continuity.Then, by one The width of pixel is added with recorded successional characteristic quantity, so that the value of successional characteristic quantity is updated into two pictures The width of element.Repeat this calculating, and continue to calculate, until the poor absolute value between pixel value is equal to or more than picture Element value between comparison when threshold value.In case of fig. 10, the pixel value for the pixel for calculating object is used as in being calculated the 4th For 120 and 200, the poor absolute value (=80) between pixel value is more than threshold value (=20), therefore now stops calculating.With three The corresponding value of the width of pixel is stored in the successional characteristic quantity in concerned pixel as pixel value.Note, continuity Width unit can be " pixel " or " mm " etc. actual size.In addition, Figure 10 shows that the positive side in x directions is searched The successional example of rope.However, it is possible to search in minus side or on both sides continuity.By entering for all pixels in image The above-mentioned calculating of row, can calculate the successional characteristic quantity for representing pixel value for each pixel.

Note, the continuity of pixel value is determined using the poor absolute value between neighborhood pixels.However, it is possible to use its His index determines continuity.For example, whether can come according to the ratio between the pixel value of neighborhood pixels equal to or less than threshold value Determine continuity.Alternatively, the pixel values of the multiple pixels that can be included based on predefined paths determines continuity.Example Such as, when it is determined that in Fig. 9 from pixel (2,3) to the continuity of pixel (5,3) when, can use pixel (2,3), (3,3), (4, 3) with the variance of the pixel value of (5,3).In this case, can if the value of the variance of pixel value is less than given threshold value Determine that pixel has continuity.

In addition, determining condition as another continuity, it may be considered that following condition：It is equal to or more than give when reaching to have When determining the pixel of pixel value (or the pixel value equal to or less than given threshold value or pixel value less than given threshold value) of threshold value, Stop successional calculating.Especially, obtained if reached as the pixel value for the pixel for calculating object in step S1050 , user concern pixel value scope outside value, then can stop calculate.These continuitys need not be used alone true Each in fixed condition, and the combination of multiple conditions can be used.

It can be any road in three dimensions for the predefined paths that calculate the successional characteristic quantity for representing pixel value Footpath, as long as the path passes through concerned pixel.If in addition, using multiple predefined paths, will can make a reservation for for each The characteristic quantity of path computing is set to multivalue vector data, and remains the characteristic quantity of concerned pixel.Alternatively, Ke Yitong Cross and calculate individually calculated using multiple predefined paths successional characteristic quantity and to obtain the value of a characteristic quantity, and It is set to the characteristic quantity of concerned pixel.The representative illustration of predefined paths is the example shown in Figure 11 A to Figure 11 D.By In space limitation, Figure 11 A to Figure 11 D show following example：It is neighbouring by 8 on two dimensional surface including concerned pixel Path in the positive direction of pixel and the four direction of negative direction is set to predefined paths.In the present embodiment, it is possible to using such as Under method：Extension and this method identical method in three dimensions, and by the positive direction including 26 neighborhood pixels and 13 directions of negative direction are set to predefined paths.Note, edge that can be from concerned pixel to image sets predefined paths.Such as The body region of object in fruit image is known, then expects only to set predefined paths in body region.Alternatively, may be used Using by the length limitation of predefined paths as the length shorter than predetermined length, and can be without for equal or longer than pre- fixed length The calculating of the length of degree.

When pixel value along predefined paths equably in the presence of, represent the output of successional characteristic quantity of pixel value often It is big.Therefore, it is if there is the block region (such as liver or heart) formed by the pixel groups with similar pixel value, then special Become in each in multiple paths of the output valve for the amount of levying big.On the contrary, for the region of the narrow shape with rib etc., output Value becomes big only in the path along shape, and diminishes in each of residual paths.By using this point, using multiple Predefined paths represent the successional characteristic quantity of pixel value for each path computing, and use result in combination, so that The block region (region in addition to bone region) of liver etc. can be strengthened by obtaining.As by for the characteristic quantity in multiple paths The method that is combined of result of calculation, can have acquisition average value, intermediate value, maximum, minimum value, minimum value and maximum it Between difference ratio or variance yields method.

Note, the method that the characteristic quantity calculates concerned pixel based on multiple paths is in the likelihood score in the region to be extracted (method in enhancing block region) is not limited to the above method, and the technology of machine learning etc. can be used to be formed.For example, In the case of the study in the region for having been defined for being extracted, created by learning for calculating the feature for each path Measure and feature based amount determines the discriminator whether concerned pixel is in the region to be extracted.At this point it is possible to which characteristic quantity is turned Change invariable rotary characteristic quantity into, and can be learnt based on the value after conversion.It is then possible to use created discriminator To extract the region of image.Note, the example of the computational methods for characteristic quantity to be converted into invariable rotary characteristic quantity is as follows Method：One in the maximum all paths of characteristic quantity is set to X-direction, and the path orthogonal with the path is worked as In, the path that characteristic quantity is maximum be set to Y direction.

In step S205, step is extracted in by the known extracting method of threshold process or pattern cut segmentation etc. Enhanced piece of region (region in addition to bone region) in rapid S204.Note, in the extraction process, in each pixel, In addition to the characteristic quantity for strengthening block region, pixel value itself or other characteristic quantities can also be used in combination.

Using above-mentioned processing, in each pixel in the picture, using multiple predefined paths by pixel, based on each The similarity between neighborhood pixels value on path, to calculate the successional characteristic quantity for representing pixel value, and feature based Measure and extract region, be enable to extract include with pixel value in bone region it is close it is pixel value, except bone region Block region in addition.That is, by using the above method for each representing successional characteristic quantity, it can solve the problem that the pixel value of bone connects The pixel value of the nearly organ that experienced radiography shooting, and therefore, it is difficult to distinguished based on the pixel value between bone and organ Problem.

(step S1060：Change the pixel value in difference image)

In step S1060, pixel value, which changes unit 53, to assign probability 0 to the region extracted in step S1055, And the region of the region exterior to being extracted assigns probability 1, and weight coefficient W is used as thereby using probability.Alternatively, except Beyond the weight coefficient W of first embodiment, the extraction result of pixel outside region-of-interest can also be used as second Weight coefficient W2.That is, pixel value change unit 53 can by using the pixel p in difference image TS difference value TS (p), plus Weight coefficient W (p) and the second weight coefficient W2 (p) calculates TS (p) × W (p) × W2 (p), poor to use multiple weight coefficients to change Pixel value in partial image.The combination of multiple weight coefficients can be such as TS (p) × (W (p)+W2 (p)) linear combination.

Note, the situation for extracting the pixel outside region-of-interest is described above.However, it is possible to according to representing pixel The value of the successional characteristic quantity of value obtains weight coefficient W2.It is set to by bone region in the present embodiment of region-of-interest, it is excellent Selection of land is configured so that bigger with the value of characteristic quantity, then the weight coefficient that the pixel value in difference image is multiplied becomes to get over Small (close to 0), and it is smaller with the value of characteristic quantity, then and weight coefficient becomes bigger (close to 1).Note, in this case, The extracted region processing in step S205 can be skipped.By setting weight coefficient by this way, it can be anticipated that with as follows Effect：When noting the difference value of bone, relatively in the block region of heart or liver outside reduction region-of-interest etc. Difference value.

In the image processing techniques according to the present embodiment, it can be adjusted based on the information in the region being not concerned with user The weight of pixel value on whole difference image.The processing can suppress retouching for the change on the position in the region that user is not concerned with Paint, so as to improve the observability of the change on the position of user's concern.

Notice that the characteristic quantity calculating method in the block region according to described in the step S1055 of above-described embodiment can be used in Purpose in addition to improving the observability of the region-of-interest on difference image.For example, in order to extract/medical imaging analysis it is all The another object of the organic region of such as heart or liver, can use the characteristic quantity for representing the likelihood score as organic region.

Other embodiment

(it also can also more completely be referred to as that " non-transitory computer can by reading and performing record in storage medium Read storage medium ") on computer executable instructions (for example, one or more programs) to perform one in above-described embodiment Individual or more function and/or including for performing one of one or more functions in above-described embodiment Or more circuit (for example, application specific integrated circuit (ASIC)) system or the computer of device, to realize the implementation of the present invention Example, and it is possible to can using for example being read by the computer by system or device and performing the computer from storage medium Execute instruction is to perform one or more function in above-described embodiment and/or control one or more circuits Method to perform one or more functions in above-described embodiment, to realize embodiments of the invention.Computer can be with Including one or more processors (for example, CPU (CPU), microprocessing unit (MPU)), and it can include dividing The computer or the network of separated processor opened, to read and perform computer executable instructions.Computer executable instructions For example computer can be provided to from network or storage medium.Storage medium can include such as hard disk, random access memory Device (RAM), read-only storage (ROM), the memory of distributed computing system, CD (such as compact disk (CD), digital universal CD (DVD) or Blu-ray Disc (BD)^TM), it is one or more in flash memory device and storage card etc..

Embodiments of the invention can also be realized by following method, i.e. pass through network or various storage mediums The software (program) of function that above-described embodiment will be performed is supplied to system or device, the computer of the system or device or in The method that Central Processing Unit (CPU), microprocessing unit (MPU) read simultaneously configuration processor.

Although with reference to exemplary embodiment, invention has been described, but it is to be understood that public the invention is not restricted to institute The exemplary embodiment opened.Most wide explanation should be given to scope of the following claims, so that it covers all these changes Type example and equivalent 26S Proteasome Structure and Function.

Claims

1. a kind of image processing apparatus, described image processing unit includes：

Obtaining unit, it is configured to obtain the first image of subject and the second image of the subject；

Difference unit, it is configured to obtain the difference image after by described first image and second image registration； And

Change unit, it is configured to based on using the pixel value in the pixel value and second image in described first image The likelihood score calculated, the processing for the pixel value being changed in the difference image.

2. image processing apparatus according to claim 1, wherein, the change unit is based on using in described first image Pixel value and pay close attention to pixel value in region distribution the likelihood score that is calculated of distributed intelligence and using described the Pixel value in two images and pay close attention to pixel value in region distribution the likelihood score that is calculated of distributed intelligence in compared with Big person, come the processing of pixel value being changed in the difference image.

3. image processing apparatus according to claim 1, wherein, the change unit is based on using in described first image Pixel pixel value and the pixel neighborhood pixels pixel value and the pixel in second image pixel value and should The likelihood score that the pixel value of the neighborhood pixels of pixel is calculated, come the processing of pixel value being changed in the difference image.

4. image processing apparatus according to claim 1, wherein, the unit that changes using following item based on being calculated Likelihood score, come the processing of pixel value being changed in the difference image：Pay close attention to the distribution of pixel value in region Distributed intelligence and pixel value in described first image and the pixel value in second image pair and the picture that obtains Element value.

5. image processing apparatus according to claim 1, described image processing unit also includes：

Deformation unit, it is configured to by the position based on each pixel in described first image and second image Each pixel position between corresponding relation make second anamorphose so that each pixel in second image Matched with the respective pixel in described first image, by described first image and second image registration.

6. image processing apparatus according to claim 5, wherein, the difference unit obtains correspondence from the image after registration Pixel value at position, and obtain difference image by carrying out difference processing for the pixel value obtained.

7. image processing apparatus according to claim 1, described image processing unit also includes：

Display processing unit, it is configured to show described first image and second image on the display unit,

Wherein, for each pixel in the difference image, the display condition of the change unit based on the display unit, Obtain the pixel value information of the scope on the pixel value in region-of-interest.

8. image processing apparatus according to claim 7, wherein, the change unit is based on described first image and described The display condition of at least one of second image, obtains the pixel value information of the scope on the pixel value in region-of-interest.

9. image processing apparatus according to claim 1, wherein, it is described for each pixel in the difference image Change unit based on the information inputted via operating unit, obtain the pixel value letter of the scope on the pixel value in region-of-interest Breath.

10. image processing apparatus according to claim 7, wherein, the change unit is based on the pixel value information, changes Become the distributed intelligence of the distribution for the pixel value paid close attention in region.

11. image processing apparatus according to claim 10, described image processing unit also includes：

Memory cell, it is configured to the distributed intelligence of the distribution of the pixel value in storage expression region-of-interest,

Wherein, the memory cell stores the multiple distributed intelligences corresponding with the region-of-interest of the subject, and

The change unit is according to the distributed intelligence obtained based on the pixel value information from the memory cell, to set point Cloth information.

12. image processing apparatus according to claim 1, wherein, the change unit, which is used, is based on the likelihood score institute The weight coefficient of acquisition, to change the pixel value in the difference image.

13. a kind of image processing apparatus, described image processing unit includes：

Change unit, it is configured in the pixel value based on each pixel in described first image and second image Comparison between the pixel value of each pixel, the processing for the pixel value being changed in the difference image.

14. image processing apparatus according to claim 13, wherein, the change unit is used to be obtained based on described compare The weight coefficient obtained, to change the pixel value in the difference image.

15. a kind of image processing apparatus, described image processing unit includes：

Difference unit, it is configured to obtain the difference image after by described first image and second image registration；

Display processing unit, it is configured to show described first image and second image on the display unit；And

Change unit, it is configured to the display condition based on the display unit, is changed the picture in the difference image The processing of element value.

16. image processing apparatus according to claim 15, wherein, the change unit, which is used, is based on the display condition And the weight coefficient obtained, to change the pixel value in the difference image.

17. image processing apparatus according to claim 16, wherein

The display condition includes, represent pixel value range intermediate value arranges value and represent relative to intermediate value pixel value model The arranges value of the width enclosed, and

It is described to change the width that unit uses arranges value and expression pixel value range based on the intermediate value for representing pixel value range Arranges value and the weight coefficient that obtains, come the processing of pixel value being changed in the difference image.

18. image processing apparatus according to claim 1, wherein, described first image and second image are with not The image obtained with timing.

19. a kind of image processing method, described image processing method includes：

Obtain the first image of subject and the second image of the subject；

Obtain the difference image after by described first image and second image registration；And

Based on the likelihood score calculated using the pixel value in the pixel value in described first image and second image, carry out Change the processing of the pixel value in the difference image.

20. a kind of image processing apparatus, described image processing unit includes：

Obtaining unit, it is configured to the image for obtaining subject；And

Characteristic quantity obtaining unit, it is configured to, and for the concerned pixel in described image, sets by the concerned pixel Multiple predefined paths, for each in path, based on the similarity between the neighborhood pixels value on path, calculate delegated path On pixel value successional evaluation of estimate, and based on the evaluation of estimate obtained for each path, obtain the concerned pixel Characteristic quantity.

21. image processing apparatus according to claim 20, wherein, the path is arranged to include each of described image Axial direction.

22. image processing apparatus according to claim 20, wherein, the characteristic quantity obtaining unit is also based on the feature Amount calculates the likelihood score that the concerned pixel is in block region.

23. image processing apparatus according to claim 20, described image processing unit also includes：

Extraction unit, it is configured to extract region from described image based on the characteristic quantity.

24. a kind of computer-readable recording medium for the computer program that is stored with, the computer program perform claim requirement 19 The step of described image processing method.